refactor: move legacy APIs to a separate module (#5381)

* refactor: move legacy APIs to a separate module

Signed-off-by: Frost Ming <me@frostming.com>

Author: frostming

docs/README.md

@@ -213,7 +213,7 @@ extension to include documentation from docstring. For example, a `.rst` documen
 create a section made from a Python Class's docstring, using the following syntax:
 
 ```rst
-.. autoclass:: bentoml.Service
+.. autoclass:: bentoml.Bento
     :members: api
 ```
 

docs/source/build-with-bentoml/snippets/metrics/runner_impl.py

@@ -1,11 +1,14 @@
-class NLTKSentimentAnalysisRunnable(bentoml.Runnable):
+import bentoml
+
+
+class NLTKSentimentAnalysisRunnable(bentoml.legacy.Runnable):
     SUPPORTED_RESOURCES = ("cpu",)
     SUPPORTS_CPU_MULTI_THREADING = False
 
     def __init__(self):
         self.sia = SentimentIntensityAnalyzer()
 
-    @bentoml.Runnable.method(batchable=False)
+    @bentoml.legacy.Runnable.method(batchable=False)
     def is_positive(self, input_text: str) -> bool:
         start = time.perf_counter()
         scores = [

src/_bentoml_impl/frameworks/catboost.py

@@ -200,12 +200,12 @@ def save_model(
 
 
 @deprecated(suggestion="Use `get_service` instead.")
-def get_runnable(bento_model: bentoml.Model) -> t.Type[bentoml.Runnable]:
+def get_runnable(bento_model: bentoml.Model) -> t.Type[bentoml.legacy.Runnable]:
     """
     Private API: use :obj:`~bentoml.Model.to_runnable` instead.
     """
 
-    class CatBoostRunnable(bentoml.Runnable):
+    class CatBoostRunnable(bentoml.legacy.Runnable):
         SUPPORTED_RESOURCES = ("nvidia.com/gpu", "cpu")
         SUPPORTS_CPU_MULTI_THREADING = True
 

src/_bentoml_impl/frameworks/lightgbm.py

@@ -204,12 +204,12 @@ def save_model(
 
 
 @deprecated(suggestion="Use `get_service` instead.")
-def get_runnable(bento_model: bentoml.Model) -> t.Type[bentoml.Runnable]:
+def get_runnable(bento_model: bentoml.Model) -> t.Type[bentoml.legacy.Runnable]:
     """
     Private API: use :obj:`~bentoml.Model.to_runnable` instead.
     """
 
-    class LightGBMRunnable(bentoml.Runnable):
+    class LightGBMRunnable(bentoml.legacy.Runnable):
         # LightGBM only supports GPU during training, not for inference.
         SUPPORTED_RESOURCES = ("cpu",)
         SUPPORTS_CPU_MULTI_THREADING = True

src/_bentoml_impl/frameworks/mlflow.py

@@ -213,14 +213,14 @@ def import_model(
 
 
 @deprecated(suggestion="Use `get_service` instead.")
-def get_runnable(bento_model: bentoml.Model) -> t.Type[bentoml.Runnable]:
+def get_runnable(bento_model: bentoml.Model) -> t.Type[bentoml.legacy.Runnable]:
     """
     Private API: use :obj:`~bentoml.Model.to_runnable` instead.
     """
     assert "predict" in bento_model.info.signatures
     predict_signature = bento_model.info.signatures["predict"]
 
-    class MLflowPyfuncRunnable(bentoml.Runnable):
+    class MLflowPyfuncRunnable(bentoml.legacy.Runnable):
         # The only case that multi-threading may not be supported is when user define a
         # custom python_function MLflow model with pure python code, but there's no way
         # of telling that from the MLflow model metadata. It should be a very rare case,
@@ -233,7 +233,7 @@ def __init__(self):
             super().__init__()
             self.model = load_model(bento_model)
 
-        @bentoml.Runnable.method(
+        @bentoml.legacy.Runnable.method(
             batchable=predict_signature.batchable,
             batch_dim=predict_signature.batch_dim,
             input_spec=None,

src/_bentoml_impl/frameworks/sklearn.py

@@ -92,7 +92,7 @@ def save_model(
         name: Name for given model instance. This should pass Python identifier check.
         model: Instance of model to be saved.
         signatures: Methods to expose for running inference on the target model. Signatures are
-                    used for creating Runner instances when serving model with bentoml.Service
+                    used for creating Runner instances when serving model with bentoml.legacy.Service
         labels: user-defined labels for managing models, e.g. team=nlp, stage=dev
         custom_objects: user-defined additional python objects to be saved alongside the model,
                         e.g. a tokenizer instance, preprocessor function, model configuration json
@@ -164,7 +164,7 @@ def get_runnable(bento_model: Model):
     Private API: use :obj:`~bentoml.Model.to_runnable` instead.
     """
 
-    class SklearnRunnable(bentoml.Runnable):
+    class SklearnRunnable(bentoml.legacy.Runnable):
         SUPPORTED_RESOURCES = ("cpu",)
         SUPPORTS_CPU_MULTI_THREADING = True
 

src/_bentoml_impl/frameworks/xgboost.py

@@ -217,12 +217,12 @@ def save_model(
 
 
 @deprecated(suggestion="Use `get_service` instead.")
-def get_runnable(bento_model: bentoml.Model) -> t.Type[bentoml.Runnable]:
+def get_runnable(bento_model: bentoml.Model) -> t.Type[bentoml.legacy.Runnable]:
     """
     Private API: use :obj:`~bentoml.Model.to_runnable` instead.
     """
 
-    class XGBoostRunnable(bentoml.Runnable):
+    class XGBoostRunnable(bentoml.legacy.Runnable):
         SUPPORTED_RESOURCES = ("nvidia.com/gpu", "cpu")
         SUPPORTS_CPU_MULTI_THREADING = True
 

src/_bentoml_sdk/service/factory.py

@@ -18,7 +18,6 @@
 from starlette.applications import Starlette
 from typing_extensions import Unpack
 
-from bentoml import Runner
 from bentoml._internal.bento.bento import Bento
 from bentoml._internal.bento.build_config import BentoEnvSchema
 from bentoml._internal.configuration.containers import BentoMLContainer
@@ -28,6 +27,7 @@
 from bentoml._internal.utils import dict_filter_none
 from bentoml.exceptions import BentoMLConfigException
 from bentoml.exceptions import BentoMLException
+from bentoml.legacy import Runner
 
 from ..images import Image
 from ..method import APIMethod

src/bentoml/__init__.py

@@ -32,11 +32,6 @@
     "server_context": "._internal.context:server_context",
     "Model": "._internal.models:Model",
     "monitor": "._internal.monitoring:monitor",
-    "Resource": "._internal.resource:Resource",
-    "Runnable": "._internal.runner:Runnable",
-    "Runner": "._internal.runner:Runner",
-    "Strategy": "._internal.runner.strategy:Strategy",
-    "Service": "._internal.service:Service",
     "Tag": "._internal.tag:Tag",
     "load": "._internal.service.loader:load",
     "Cookie": "._internal.utils.http:Cookie",
@@ -50,9 +45,6 @@
     "pull": ".bentos:pull",
     "push": ".bentos:push",
     "serve": ".bentos:serve",
-    # Legacy APIs
-    "HTTPServer": ".server:HTTPServer",
-    "GrpcServer": ".server:GrpcServer",
     # New SDK
     "service": "_bentoml_sdk:service",
     "runner_service": "_bentoml_sdk:runner_service",
@@ -83,6 +75,7 @@
     from _bentoml_impl.frameworks import xgboost
 
     from . import bentos
+    from . import legacy
 
     # BentoML built-in types
     from ._internal.bento import Bento
@@ -94,11 +87,6 @@
     from ._internal.context import server_context
     from ._internal.models import Model
     from ._internal.monitoring import monitor
-    from ._internal.resource import Resource
-    from ._internal.runner import Runnable
-    from ._internal.runner import Runner
-    from ._internal.runner.strategy import Strategy
-    from ._internal.service import Service
     from ._internal.service.loader import load
     from ._internal.tag import Tag
     from ._internal.utils.args import use_arguments
@@ -176,6 +164,7 @@
     FrameworkImporter.install()
 
     bentos = _LazyLoader("bentoml.bentos", globals(), "bentoml.bentos")
+    legacy = _LazyLoader("bentoml.legacy", globals(), "bentoml.legacy")
 
     # ML Frameworks
     catboost = _LazyLoader(
@@ -301,22 +290,33 @@
     del _LazyLoader, FrameworkImporter
 
     def __getattr__(name: str) -> Any:
+        import bentoml.legacy as legacy
+
         if name in MODULE_ATTRS:
             from importlib import import_module
 
             module_name, attr_name = MODULE_ATTRS[name].split(":")
             module = import_module(module_name, __package__)
             return getattr(module, attr_name)
-        raise AttributeError(f"module {__name__} has no attribute {name}")
+        elif name in legacy.__all__:
+            from ._internal.utils import warn_deprecated
+
+            warn_deprecated(
+                f"`bentoml.{name}` is moved to `bentoml.legacy.{name}` "
+                "and will be removed in a future version.",
+            )
+            return getattr(legacy, name)
+        else:
+            raise AttributeError(f"module {__name__} has no attribute {name}")
 
 
 __all__ = [
     "__version__",
     "Context",
     "Cookie",
-    "Service",
     "bentos",
     "models",
+    "legacy",
     "batch",
     "metrics",
     "container",
@@ -326,8 +326,6 @@ def __getattr__(name: str) -> Any:
     "io",
     "Tag",
     "Model",
-    "Runner",
-    "Runnable",
     "monitoring",
     "BentoCloudClient",  # BentoCloud REST API Client
     # bento APIs
@@ -374,8 +372,6 @@ def __getattr__(name: str) -> Any:
     "load_config",
     "save_config",
     "set_serialization_strategy",
-    "Strategy",
-    "Resource",
     # new SDK
     "service",
     "runner_service",

src/bentoml/_internal/frameworks/common/pytorch.py

@@ -56,7 +56,7 @@ def __init__(self, *inner_args: t.Any, **inner_kwargs: t.Any) -> None:
     return NewClass
 
 
-class PytorchModelRunnable(bentoml.Runnable):
+class PytorchModelRunnable(bentoml.legacy.Runnable):
     SUPPORTED_RESOURCES = ("nvidia.com/gpu", "cpu")
     SUPPORTS_CPU_MULTI_THREADING = True